Rankings / Unknown
Qwen3 Coder 30B A3b Instruct
released 2025-07-31
BenchAtlas Index
as of 2026-07-14
base configuration
rank #377
10 families · 5 categories · high
Benchmark evidence
13 results
agentic coding
| τ²-Bench | 34.5 % | independent | Artificial Analysis |
coding
| LiveCodeBench v6 implementation=artificial-analysis | 40.3 % | independent | Artificial Analysis | |
| SciCode | 27.8 % | independent | Artificial Analysis |
external indices
| Intelligence Index v4.1 | 13.6 points | independent | Artificial Analysis | |
| AA Math Index | 29.0 points | independent | Artificial Analysis |
knowledge science
| GPQA Diamond implementation=artificial-analysis | 51.6 % | independent | Artificial Analysis | |
| Humanity's Last Exam implementation=artificial-analysis | 4.0 % | independent | Artificial Analysis | |
| MMLU-Pro implementation=artificial-analysis | 70.6 % | independent | Artificial Analysis |
long context instruction
| AA-LCR | 29.0 % | independent | Artificial Analysis | |
| IFBench | 32.6 % | independent | Artificial Analysis |
reasoning math
| AIME implementation=artificial-analysis | 29.7 % | independent | Artificial Analysis | |
| AIME year=2025 · implementation=artificial-analysis | 29.0 % | independent | Artificial Analysis | |
| MATH-500 implementation=artificial-analysis | 89.3 % | independent | Artificial Analysis |
Agent + model results
systems, not bare-model scores
| agent + model EntroPO + Qwen3 Coder 30B A3b Instruct | SWE-bench Verified | 60.4 % | unverified | SWE-bench Leaderboard |
| agent + model EntroPO + Qwen3 Coder 30B A3b Instruct | SWE-bench Verified | 52.2 % | unverified | SWE-bench Leaderboard |
| agent + model OpenHands + Qwen3 Coder 30B A3b Instruct | SWE-bench Verified | 51.6 % | unverified | SWE-bench Leaderboard |
| agent + model EntroPO + Qwen3 Coder 30B A3b Instruct | SWE-bench Lite | 49.7 % | unverified | SWE-bench Leaderboard |
| agent + model EntroPO + Qwen3 Coder 30B A3b Instruct | SWE-bench Lite | 45.0 % | unverified | SWE-bench Leaderboard |
| agent + model Artificial Analysis harness + Qwen3 Coder 30B A3b Instruct | Terminal-Bench Hard | 15.2 % | independent | Artificial Analysis |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.
